Microservices and modularity with
Java
Elek Márton
@anzix
2015 December
DPC Consulting

Agenda
● Microservice definition
● Core implementation patterns

People need modularity not
microservices

Microservices is one way to achieve
modularity.
Other ways: OSGi, SPI, java9 jigsaw

What are microservices?
The fire making rule:
● Use more sticks than you think they would be
enough
● Use a little bit smaller sticks than you think would
be enough

Informal definition
Microservices when you have smaller, separated
services as usual
● Many times they are not micro (group of services)
● Many times REST is used (but not always)
● Separated means different process/JVM instances

More formal definition
● “an approach to developing a single application as
a suite of small services,
● each running in its own process
● and communicating with lightweight mechanisms,
often an HTTP resource API.
These services are built around business capabilities
and independently deployable by fully automated
deployment machinery.”
Lewis, Fowler

Microservices and REST
REST is just one solution to communicate
● between services
● or call any of the service

Microservices vs SOA
Even if the “Microservices != SOA” a very common
statement there are a lot of shared concept.
● Usually we have new tools (using existing solutions
differently) not totally new concept

Implementation patterns

Monolith application

Spring boot example
@EnableAutoConfiguration
@RestController
@ComponentScan
public class TimeStarter {
@Autowired
TimeService timerService;
@RequestMapping("/now")
public Date now() {
return timerService.now();
}
public static void main(String[] args) {
SpringApplication.run(TimeStarter.class, args);
}
}

Modular monolith

API gateway

API gateway
Goals:
● Hide available microservices behind a service
facade pattern
– Routing, Authorization
– Deployment handling, Canary testing
– Logging, SLA
Implementations:
● Spring cloud: Netflix Zuul
● Netflix Zuul based implementation
● Twitter Finagle based implementation
● Amazon API gateway
● Simple Nginx reverse proxy configuration

Spring Cloud + Zuul example
@EnableAutoConfiguration
@ComponentScan
@EnableEurekaClient
@EnableZuulProxy
public class ApiGatewayStarter {
public static void main(String[] args) {
SpringApplication.run(ApiGatewayStarter.class, args);
}
}

Service registry

Service registry
Goal:
● Store the location and state of the available
services
– Health check
– DNS interface
Implementations:
● Spring cloud: Eureka
● Netflix eureka based implementation
● Consul.io
● etcd, zookeeper
● Simple: DNS or hosts file

Eureka

Eureka server
@SpringBootApplication
@EnableEurekaServer
public class EurekaStarter {
public static void main(String[] args) {
new SpringApplicationBuilder(EurekaStarter.class).web(true).run(args);
}
}

Eureka client
@EnableAutoConfiguration
@RestController
@ComponentScan
@EnableEurekaClient
public class TimeStarter {
@Autowired
TimeService tttService;
@RequestMapping("/now")
public Date now() {return tttService.now();}
public static void main(String[] args) {
SpringApplication.run(TimeStarter.class, args);
}
}

SIMPLE Discovery client
@Service
public class Client {
@Autowired
private DiscoveryClient discoveryClient;
public void list() {
for (ServiceInstance instance : discoveryClient.getInstances("BOOTSTRAP"))
{
System.out.println(instance.getHost());
System.out.println(instance.getPort());
System.out.println(instance.getUri());
}
}
}

Config service

Monolith configuration
● Store configuration outside of the project
– Versioning? Who did what
● Store configuration in separated versioned project
– Which binary version need which config version?
● Store configuration with the source code
– Sensitive configuration?

Config Service
Goals:
● One common place for all of the configuration
– Versioning
– Auditing
– Multiple environment support
Solutions:
● Spring Cloud config service
● Zookeeper
● Most of the service registry have key->value store
● Any persistence datastore

Config server
@SpringBootApplication
@EnableConfigServer
@EnableEurekaClient
public class ConfigServer {
public static void main(String[] args) {
new SpringApplicationBuilder(ConfigServer.class).run(args);
}
}

Config repository
elek@sc /tmp/config-repo [master] > ls -lah
drwxr-xr-x 3 elek elek 100 Dec 14 14:32 .
drwxrwxrwt 57 root root 1.8K Dec 14 14:37 ..
-rw-r--r-- 1 elek elek 55 Dec 14 14:16 application.properties
drwxr-xr-x 8 elek elek 260 Dec 14 14:40 .git
-rw-r--r-- 1 elek elek 12 Dec 14 14:32 timer.properties

Config client
@Service
public class TimeService {
@Value("${version:Unknown}")
String version;
public Date now() {
return new Date();
}
public String version() {
return version;
}
}

Monitoring

Monitoring
Goal:
● Combined business log
● Health check, exceptions, what is running?
● Metrics
Solutions (for one of the goals):
● Netflix Spectator, Servo, Atlas
● Spring cloud: Hystrix (circuit breaker)
● ELK: ElasticSearch + Logstash + Kibana
● Cloud based solutions (logentries...)
● Metrics: Dropwizard based implementations

Kibana (from ELK)

Hystrix

Hystrix
@Service
public class TimeService {
@HystrixCommand(fallbackMethod = "safeNow")
public Date now() {
return new Date();
}
public Date safeNow() {
return new Date(0);
}
}

Inter-module calls

Inter-module calls

Inter-module calls
Goal:
● RPC calls between modules
● Protocol
● Usage of the Service Registry
Solutions:
● Spring cloud/Netflix: ribbon
● Simple REST calls
● Message Oriented solutions:
– Rabbit MQ, ZMQ, ...

Distributed tracing

Distributed tracing (simlified)

Distributed tracing
Goal:
● Global information about latency and errors
– Service enter/exit point measurement
– Transaction identification
Implementations:
● Twitter’s Zipkin
● Apache (Cloudera) HTrace
● Spring cloud Sleuth

Zipkin
source: https://bitbucket.org/aktivecortex/aktivecortex/wiki/tracing.md

Summary

Advantages of monolith
Modular monolith Microservices

Advantages of microservices
● Modular, loosely coupled
● Easily scalable to multiple machines
● Separated, modular development lifecycle
● Partial deploy without restart [like OSGi]
● Supports heterogeneous architecture (tech stack
migration)

Summary
● Focus on modularity
● Be familiar with the microservice related tools: they
could be useful even with monolith applications…

Start with monolith?
Stefan Tilkov:
http://martinfowler.com/articles/dont-
start-monolith.html
Martin Fowler:
http://martinfowler.com/bliki/MonolithFirst.
html

Q&A
Márton Elek
DPC consulting http://dpc.hu
twitter: @anzix
http://blog.dpc.hu, http://blog.anzix.net